The core idea: one asset, one price
Arbitrage exploits situations where the same economic exposure trades at two different prices. Buy the cheap leg, sell the expensive leg, and wait for convergence. The profit is the gap minus costs. In efficient markets those gaps are tiny and short-lived — which is exactly why understanding the mechanics matters more than spotting the gap.
1. Pure (riskless) arbitrage
The textbook case: identical instruments, simultaneous execution.
- Exchange arbitrage: The same asset quoted at different prices on two venues. High-frequency firms with co-located servers capture these in microseconds; retail traders essentially never see them.
- Triangular FX arbitrage: Three currency pairs whose cross-rates drift out of alignment. Again, machine territory.
2. Basis and carry arbitrage
Here the two legs are *related but not identical*, so the trade carries real risk until the convergence date.
- Cash-and-carry: Buy the spot asset, sell the futures contract. The futures price must converge to spot at expiry, so the annualized basis is a quasi-interest rate. When crypto perpetual funding rates spike, this is the trade professionals run against them.
- The Treasury basis trade: Buy cash bonds, short bond futures, lever the tiny spread. Profitable and stable — until a funding shock forces mass unwinds, as happened in March 2020.
3. Statistical arbitrage
Statistical arbitrage bets on the *historical relationship* between instruments rather than a hard convergence guarantee.
- Pairs trading: Long one stock, short a highly correlated peer when their spread stretches beyond historical norms. The bet is mean reversion, and it can simply be wrong when fundamentals genuinely diverge.
- Index arbitrage: Trading a basket of stocks against the index future when the basket drifts from fair value.
Why arbitrage trades still fail
- Funding risk: Convergence trades are usually levered. If your financing is pulled before the spread closes, you realize the loss even though the thesis was right. "The market can stay irrational longer than you can stay solvent" is an arbitrage epitaph.
- Execution slippage: The gap must exceed both spreads, both commissions, and any borrow fees on the short leg. Most visible "arbitrage" disappears after honest cost accounting.
- Correlation breakdown: Statistical relationships hold until a regime change breaks them, and the losses arrive precisely when everything else is also going wrong.